Mimivirus Giant Particles Incorporate a Large Fraction of Anonymous and Unique Gene Products

Abstract

Acanthamoeba polyphaga mimivirus is the largest known virus in both particle size and genome complexity. Its 1.2-Mb genome encodes 911 proteins, among which only 298 have predicted functions. The composition of purified isolated virions was analyzed by using a combined electrophoresis/mass spectrometry approach allowing the identification of 114 proteins. Besides the expected major structural components, the viral particle packages 12 proteins unambiguously associated with transcriptional machinery, 3 proteins associated with DNA repair, and 2 topoisomerases. Other main functional categories represented in the virion include oxidative pathways and protein modification. More than half of the identified virion-associated proteins correspond to anonymous genes of unknown function, including 45 “ORFans.” As demonstrated by both Western blotting and immunogold staining, some of these “ORFans,” which lack any convincing similarity in the sequence databases, are endowed with antigenic properties. Thus, anonymous and unique genes constituting the majority of the mimivirus gene complement encode bona fide proteins that are likely to participate in well-integrated processes.

Acanthamoeba polyphaga mimivirus (mimivirus) is the largest virus isolated so far (23). Based on its highly specific characteristics, this double-stranded-DNA icosahedral virus (47) is the first member of the new Mimiviridae family (33, 43). Computational annotation of its 1.2-Mb genome (33) revealed many atypical features, including the presence of key translation enzymes, a full complement of DNA repair pathway components, and the unique presence of three different topoisomerases (of types IA, IB, and II) (2, 33). Another unique characteristic of mimivirus is the presence of nearly identical promoter sequence motifs upstream of half of its 911 protein-encoding genes (42), which are presumably associated with proteins expressed during the early or late-early phase. Only 23% of the predicted coding genes exhibit convincing homology to proteins of known function, and 39% of them do not exhibit a clear (E values, <10⁻⁵) sequence database match (33). Such coding regions without sequence similarity to other genes in databases are considered orphan open reading frames (ORFs) and termed “ORFans” (12). The origin and function of ORFan genes are still a matter of controversy, with opinions ranging from considering them pieces of junk DNA (1, 8, 40, 44) to seeing them as quickly evolving sequences encoding normally expressed functional proteins (38, 39). Recent clinical evidence raised the possibility that mimivirus might be a human pathogen causing pneumonia (4, 24, 34), as suspected when it was first isolated from a cooling tower following an outbreak of pneumonia (23).

Mass spectrometry-based analysis has recently emerged as a technique of choice to identify more comprehensively the set of viral proteins associated with viral particles (19, 29, 49). We now present the application of this technique to the largest known, and presumably most complex, viral particle, Acanthamoeba polyphaga mimivirus.

MATERIALS AND METHODS

Sample preparation for bidimensional (2D) gel electrophoresis.

Acanthamoeba polyphaga mimivirus (23) was purified through a sucrose gradient (25%) and washed twice with phosphate-buffered saline (PBS) in the presence of protease inhibitors (Complete; Roche, Mannheim, Germany). The resulting pellet was solubilized in 40 mM Tris-HCl, pH 7.5, supplemented with 2% (wt/vol) sodium dodecyl sulfate (SDS; Sigma-Aldrich) and 60 mM dithiothreitol (DTT), followed by 5 min of heating at 95°C. The insoluble fraction was removed by centrifugation (12,000 × g, 4°C, 10 min), and soluble proteins were precipitated using a PlusOne 2-D cleanup kit (Amersham Biosciences) to remove SDS. The final pellet was resuspended in solubilization buffer {7 M urea, 2 M thiourea, 4% (wt/vol) 3-[(3-cholamidopropyl)-dimethylammonio]-1-propanesulfonate (CHAPS)} and stored at −80°C until isoelectric focusing (IEF) was performed.

2D gel electrophoresis and silver staining.

Immobiline DryStrips (18 cm, pH 4 to 7 or 6 to 11; Amersham) were rehydrated overnight using 350 μl rehydration buffer (8 M urea, 2% [wt/vol] CHAPS, 60 mM DTT, 2% [vol/vol] Immobiline pH gradient (IPG) buffer [Amersham]) containing 200 μg of solubilized mimivirus proteins, and IEF was carried out according to the manufacturer's protocol (Multiphor II; Amersham). Before the second-dimension electrophoresis was performed, strips were equilibrated twice in 10 ml equilibration buffer (30% [vol/vol] glycerol, 2% [wt/vol] SDS, 6 M urea, 50 mM Tris-HCl, bromophenol blue, pH 8.8) for 15 min. This buffer was supplemented with 65 mM DTT for the first equilibration and with 100 mM iodoacetamide for the second one. The strips were then embedded in 0.5% agarose, and the proteins were resolved by 10% SDS-polyacrylamide gel electrophoresis (SDS-PAGE) (Ettan DALT; Amersham) at 5 W/gel for 30 min, followed by 4 to 5 h at 17 W/gel. Following migration, gels were stained by a method compatible with mass spectrometry (36). Spots excised from the gel were stored at −20°C until identification. For accuracy, the spectra of at least two separate samples of each protein were analyzed and compared.

In-gel digestion and MALDI-TOF mass spectrometry (MS).

Spots excised from silver-stained gels were destained and subjected to in-gel digestion with trypsin (sequencing-grade modified porcine trypsin; Promega, Madison, WI) (36). Tryptic peptides were then extracted from the gel by successive treatments with 5% formic acid and 50% acetonitrile-5% formic acid. Extracts were pooled and dried in a Speedvac evaporator. Peptides resuspended in an α-cyano-4-hydroxycinnamic acid matrix solution (prepared by diluting 3 times a saturated solution in 50% acetonitrile-0.3% trifluoroacetic acid) were then spotted on the matrix-assisted laser desorption ionization (MALDI) target. Mass analyses were performed on a MALDI-time of flight (MALDI-TOF) Brüker Ultraflex spectrometer (Brüker Daltonique, Wissembourg, France). Mass spectra were internally calibrated using autolytic peptides from trypsin. Tryptic peptide mass lists were used to identify the proteins, using Mascot software. Searches were performed against all available sequences in public databases, including those for eukaryotes (http://www.matrixscience.com).

Immunization and Western blotting.

Acanthamoeba polyphaga mimivirus proteins resolved by 2D gel electrophoresis were transferred onto nitrocellulose membranes (Semi-Phor unit; Hoefer Scientific, San Francisco, CA). Membranes were then blocked in PBS supplemented with 0.2% Tween 20 and 5% nonfat dry milk (PBS-Tween-milk) for 1.5 h before incubation with anti-Acanthamoeba polyphaga mimivirus sera. The sera were obtained from BALB/c mice immunized by three intraperitoneal injections (with 15-day intervals between injections) of 5 μg of purified viral particles resuspended in CpG as an adjuvant. After 1 h of incubation (1:6,400 dilution in PBS-Tween-milk), membranes were washed three times with PBS-Tween and probed with horseradish peroxidase-conjugated goat anti-mouse secondary antibodies (1:1,000; Amersham). Detection was achieved by chemiluminescence (ECL; Amersham).

SDS-PAGE coupled with electrospray ionization-ion-trap MS/MS analysis.

An Acanthamoeba polyphaga mimivirus sample (175 μg) prepared as described above was separated by SDS-PAGE on an 11% acrylamide gel (22), and the proteins were revealed afterwards by Coomassie blue R-250 staining. Bands of 1-mm thickness were systematically cut from the gel, resulting in 49 pieces to be analyzed. These bands were reduced and alkylated before trypsin digestion, and the resulting peptides were analyzed through liquid chromatography (LC) (Finnigan Surveyor HPLC system; Thermo Electron, San Jose, CA) coupled to an electrospray ionization-ion-trap mass spectrometer (LCQ-Deca XP; Thermofinnigan). Tryptic peptides were resuspended in 20 μl formic acid at 5% in water, and 1/20 of the whole extract obtained from each piece of gel was desalted on a C₁₈ trapping column (Zorbax 300SB-C18 [5-μm inner diameter, 5 by 0.3 mm]; Agilent) by a 20-min washing step with 0.1% formic acid. Peptide separation was achieved by a 60-min linear gradient of acetonitrile (0 to 65%) in 0.2% formic acid on a C₁₈ reverse-phase column (PicoFrit column [5-μm BioBasic C₁₈, 300-Å pore size, 75-μm inner diameter, 100-mm long, 15-μm tip]). The peptides were then ionized at a capillary temperature of 160°C with a 2-kV spray voltage. A collision energy of 35% was applied for the MS/MS scan events. The MS/MS spectrum of the three most intense peaks was obtained after each full MS scan. The dynamic exclusion features were set at a repeat count of 2 within 0.5 min, with an exclusion duration of 3 min.

Protein identification was performed using the TurboSequest algorithm (11) in the Bioworks 3.2 software package (Thermo Electron) and the Acanthamoeba polyphaga mimivirus database (33). TurboSequest automatically identifies proteins by matching a peptide's experimental MS/MS spectrum to a predicted mass spectrum for amino acid sequences within databases. Identified peptides were evaluated using the cross-correlation number (X_corr) versus the charge state. Criteria for positive identification of peptides were X_corr values of >1.8 for singly charged ions, >2.5 for doubly charged ions, and >3.5 for triply charged ions. Only the best match was considered, with two or more unique peptides identifying a protein. Four proteins with masses below 19 kDa were assigned by a single peptide after visual inspection of the MS/MS spectrum.

Protein glycosylation analysis.

A purified pellet of Acanthamoeba polyphaga mimivirus was resuspended in β-mercaptoethanol (Sigma-Aldrich) and incubated overnight at 4°C. The insoluble fraction was removed by centrifugation (12,000 × g, 4°C, 10 min). Proteins present in the supernatant were separated by SDS-PAGE on a 15% acrylamide gel (22) and visualized by Coomassie R-250 staining. Glycoproteins were revealed using a GlycoProfile III kit (Sigma-Aldrich), which allows specific and sensitive in-gel fluorescence detection of glycoproteins.

Sequence annotation and ORF functional predictions.

ORF sequences were annotated using homology searches against all available databases, as previously described (33). In addition, anonymous (homologous to proteins of unknown function) and orphan sequences were reanalyzed using the 3D structure-based homology program FUGUE (37), generating more tentative structural and/or functional assignments (indicated within parentheses in Table 1). Only assignments scored as “likely” or “certain” were reported.

TABLE 1.

Mimivirus proteins identified in this study

Protein name and category	ORFa	2D spot no.b
DNA topology and repair
Topoisomerase IB (pox-like)	MIMI_R194	NP
Topoisomerase IA (bacterial type—untwisting)	MIMI_L221	NP
DNA polymerase family X	MIMI_L318	NP
Regulator of chromosome condensation RCC1	MIMI_R345	1, 2, 3, 4, 5
DNA UV damage repair endonuclease	MIMI_L687	NP
Transcription
DNA-directed RNA polymerase subunit 5	MIMI_L235	93
DNA-directed RNA polymerase subunit 2	MIMI_L244	NP
VV D6R-like helicase	MIMI_R350	NP
DNA-directed RNA Pol2 G subunit	MIMI_L376	55
D6R-like putative early transcription factor	MIMI_L377	NP
African swine fever virus (ASFV) NP868R-like mRNA capping enzyme	MIMI_R382	NP
DNA-directed RNA polymerase subunit L	MIMI_R470	37
DNA-directed RNA polymerase subunit 1	MIMI_R501	NP
VV I8-like helicase	MIMI_L540	NP
NPH-1-like transcription termination factor	MIMI_L538	NP
TFIIB-like transcription initiation factor	MIMI_L544	NP
NPH-1-like transcription termination factor	MIMI_R563	NP
Oxidative pathways
Choline dehydrogenase	MIMI_R135	69, 70, 71, 72
Amine oxidoreductase	MIMI_R188	10
Glutaredoxin (ESV128 type)	MIMI_R195	NP
Thioredoxine domain	MIMI_R362	96
Thioredoxine domain	MIMI_R443	67
Zn-dependent alcohol dehydrogenase	MIMI_L498	NP
Putative oxidoreductase (C terminus)	MIMI_L893	NP
Putative oxidoreductase (N terminus)	MIMI_L894	NP
Thiol oxidoreductase E10R (C-C bonding)	MIMI_R596	87
Protein/lipid modification
Ubiquitin-specific protease	MIMI_L293	NP
Serine/threonine protein phosphatase	MIMI_R307	NP
Phosphoesterase	MIMI_R398	NP
F10L-like S/T protein kinase	MIMI_R400	NP
Putative S/T protein kinase	MIMI_L516	NP
Putative triacylglycerol lipase	MIMI_R526	45, 46
Prolyl 4-hydroxylase	MIMI_L593	NP
Particle structure
A16L-like virion-associated membrane protein	MIMI_L65	NP
Major core protein	MIMI_L410	11, 12
Capsid protein D13L	MIMI_L425	6, 7, 8, 9
Putative spore coat assembly factor	MIMI_L454	NP
Lipocalin-like outer membrane lipoprotein	MIMI_R877	NP
Other
Cytoskeletal protein	MIMI_L294	21, 22
RNA methyl transferase	MIMI_R383	NP
tRNA (uracil-5-)methyltransferase	MIMI_R407*	88, 89
Putative exoribonuclease	MIMI_R528	NP
Cytochrome P450	MIMI_L532*	75
Putative elongation factor EF-Tu homolog	MIMI_R553	73, 74
Proline-rich protein	MIMI_R610	59, 60
Mannose-6-phosphate isomerase	MIMI_L612	68
Phosphatidyl-ethanolamine-binding protein	MIMI_R644	NP
Chemotaxis	MIMI_R706	NP
Chemotaxis	MIMI_R721	44
ORFan	MIMI_L48*	65, 66
ORFan	MIMI_L208	NP
Unknown (WD repeat)	MIMI_L264	NP
ORFan	MIMI_L274	NP
Chilo iridescent virus 380R-like anonymous domain	MIMI_R301	NP
ORFan	MIMI_L309	NP
ORFan	MIMI_L330	83
ORFan	MIMI_L352	NP
ORFan	MIMI_L389	NP
ORFan	MIMI_L399	NP
Unknown (PBCV-1 A488R)	MIMI_L417	NP
ORFan	MIMI_L442	23, 24, 25, 26, 30, 31, 32, 33, 34, 35, 36
ORFan	MIMI_L452	NP
Unknown (ankyrin repeat)	MIMI_L484	NP
ORFan (biotinyl carrier)	MIMI_L485*	95
ORFan	MIMI_L488	NP
ORFan (coenzyme A binding)	MIMI_L492	NP
Unknown	MIMI_L515	NP
ORFan (nuclease)	MIMI_L533	NP
ORFan (leucine zipper)	MIMI_L550	97
Unknown	MIMI_L567	78, 79, 80, 81, 82
ORFan	MIMI_L585	NP
ORFan	MIMI_L591	77
ORFan (leucine zipper)	MIMI_L647	94
ORFan	MIMI_L688	52, 53
Unknown (WD repeat)	MIMI_L690	27, 28, 29
Unknown (COG 3979)	MIMI_L701	NP
ORFan (fusion protein [Nipah virus])	MIMI_L724	85, 86
ORFan	MIMI_L725	47, 48, 49, 50, 51
ORFan	MIMI_L778	NP
Unknown	MIMI_L829	13, 14, 15, 16, 17, 18, 54, 62
ORFan	MIMI_L851	NP
ORFan (aromatic amino acid hydrolase)	MIMI_L872	43
ORFan	MIMI_L899	NP
ORFan	MIMI_R160	NP
Unknown	MIMI_R161	NP
Unknown	MIMI_R253	38, 39, 40, 76
ORFan	MIMI_R326	NP
Unknown	MIMI_R327	NP
Unknown (ASFV C475L)	MIMI_R341	41
ORFan	MIMI_R347	NP
Unknown (ASFV S273R)	MIMI_R355	NP
ORFan (apolipophorin III)	MIMI_R387	42
Unknown	MIMI_R402	19
ORFan (kinase)	MIMI_R403	NP
ORFan	MIMI_R457	NP
ORFan	MIMI_R459	91, 92
ORFan	MIMI_R463	NP
Unknown	MIMI_R472	NP
Unknown	MIMI_R486	NP
Unknown (PBCV-1 A352L)	MIMI_R489	56, 57, 58
ORFan	MIMI_R557	NP
ORFan (acetyltransferase)	MIMI_R584	63, 64, 84
ORFan	MIMI_R646	NP
Unknown (WD repeat)	MIMI_R648	20
ORFan	MIMI_R653	61
ORFan	MIMI_R658	NP
ORFan	MIMI_R679	NP
ORFan	MIMI_R691	NP
ORFan	MIMI_R692	90
ORFan	MIMI_R695	NP
ORFan	MIMI_R705	NP
ORFan (alpha-synuclein)	MIMI_R710	NP
Unknown	MIMI_R722	NP
ORFan	MIMI_R727	NP

^a*, ORF products identified solely through 2D MS.

^bNP, proteins not positioned on 2D gels.

Immunogold labeling.

Immunolabeling of ultrathin sections for electron microscopy was performed as previously described (24). Briefly, ultrathin sections (70 nm) were cut using an ultramicrotome (Ultracut E; Leica), collected on 300-mesh nickel grids without Formvar (Electron Microscopy Sciences), and then incubated in PBS supplemented with 0.2% bovine serum albumin (Roche Diagnostics). The sections were then treated with 0.05 M lysine (Sigma-Aldrich) in order to neutralize putative residual aldehyde groups. After repeated washings with PBS, the sections were then incubated for 3 h at 37°C under a humidified atmosphere with a monoclonal anti-Mimi_L725 antibody (Mab6E2; 1:100) (23) diluted in PBS supplemented with 3% nonfat dry milk. The secondary reaction used goat anti-mouse antibodies conjugated to 25-nm colloidal gold particles (Aurion EM Reagents, The Netherlands), followed by washings with PBS and H₂O. Gold particles were visualized using an R-GENT SE-EM silver enhancement kit (Aurion) according to the instructions of the manufacturer. After washings and aldehyde fixation, a treatment with 5% uranyl acetate (ICN Biomedicals) was done to produce intense electron-opaque staining, and samples were viewed with a Philips electron microscope (MORGAGNI 268D) at 80 kV. A negative control experiment was performed with preimmune serum.

RESULTS

Global proteomic analysis of mimivirus particle by LC-MS/MS.

A critical parameter for every proteome project is the purity of the sample to be analyzed. This is particularly true when considering intracellular pathogens. For this study, the absence of major eukaryotic debris was assessed by negative-staining electron microscopy (Fig. 1A). Mimivirus proteins separated by one-dimensional SDS-PAGE (Fig. 1B) were fractionated in 49 pieces of gel, and corresponding tryptic digests were analyzed by capillary LC-MS/MS. Alternatively, mimivirus samples were resolved by 2D gel electrophoresis (Fig. 2), and protein identification was achieved by MALDI-TOF-MS. These complementary approaches allowed 114 proteins to be identified. These are listed in Table 1, together with their predicted functions (where applicable), and are classified into broad functional categories. The diversity of functions represented in the viral particle is quite large, with bona fide “structural” proteins amounting for a small fraction of the number of proteins. We considered only five candidates, whose sole role is to constitute the virion particle, as belonging to such a class. They are an A16L-like virion-associated membrane protein, the major core protein, the capsid protein D13L, a putative spore coat assembly factor, and a lipocalin-like outer membrane lipoprotein. Transcriptional enzymes and factors (12 gene products) constitute the largest functional category associated with the viral particle. This set includes all five DNA-directed RNA polymerase subunits as well as two helicases, one mRNA guanylyltransferase, and four transcription factors. Incidentally, one of the best identification scores was obtained with the β-subunit of the RNA polymerase (MIMI_R501), where the intron computationally identified within the gene was found to be excised as predicted (33). The next largest functional group is constituted of nine gene products, all of which are associated with oxidative pathways. Some of these enzymes might help the virus to cope with the oxidative stress generated by the host defense. The protein/lipid modification functional category is also well represented, including a phosphoesterase and a lipase, which are eventually used to digest the cell membrane, two protein kinases, and a protein phosphatase. Finally, five proteins associated with DNA topology and damage repair were identified, including the topoisomerase IA (MIMI_L221) and a DNA UV damage repair endonuclease (MIMI_L687), which were never described before as viral proteins. Other identified proteins belonging to this category are a tRNA methyltransferase and a putative mannose-6-phosphate isomerase (see Table S1 in the supplemental material). Yet the most abundant class of proteins associated with the mimivirus particle corresponds to those with unknown function (65 proteins), among which are 45 ORFans exhibiting no convincing sequence match in the databases. We noticed that the fraction of ORFan gene products found to be associated with the virion (45/114 [39.47%]) is equivalent to their proportion within the whole mimivirus genome (39%), a finding consistent with the notion that these genes are in no way different from other mimivirus genes and probably all encode bona fide proteins of importance to virus physiology. In a landmark study, Iyer et al. (16) identified a set of nucleocytoplasmic large DNA virus core genes that they further classified into four classes according to their decreasing levels of conservation across the main nucleocytoplasmic large DNA virus clades. We found that mimivirus particles incorporated four of the nine class I core gene products (including the major capsid protein), none of the class II core gene products, five of the class III gene products, and three of the class IV gene products (see Table S1 in the supplemental material). Finally, a small fraction (13/114) of the products originated from genes preceded by the highly conserved predicted mimivirus promoter (found in front of about half of all mimivirus genes) (42), confirming its putative role in governing the transcription of early or late-early genes (see Table S1 in the supplemental material).

FIG. 1.

Electron micrograph and SDS-PAGE of mimivirus. (A) Purified mimivirus particles (bar, 500 nm) separated by 11% SDS-PAGE and stained with Coomassie blue R250. (B) Bands in gels were excised for subsequent LC-MS/MS analysis after in-gel trypsin digestion. Identified proteins are listed in Table 1 (also see details in Table S1 in the supplemental material).

FIG. 2.

2D gel electrophoresis patterns of proteins. Acanthamoeba polyphaga mimivirus was separated using 18-cm pI 4 to 7 (left) or 6 to 11 (right) IPG strips for the first dimension (IEF) followed by 10% linear SDS-PAGE for the second dimension. Spots revealed by silver staining were cut out from the gel and subjected to trypsin digestion followed by MALDI-TOF-MS analysis. Molecular sizes and pI ranges are indicated. Identified proteins are listed in Table 1 (also see details in Table S1 in the supplemental material).

Posttranslational modifications.

As shown on 2D gels, most of the mimivirus proteins were not resolved into single spots but rather as a train of spots (Fig. 2). This was the case, for instance, for the capsid protein D13L (spots 6 to 9), which is glycosylated (Fig. 3). We observed that 66% of ORFan-encoded proteins exhibited molecular weights corresponding to their in silico predictions. Others, such as the MIMI_L442 ORFan gene product, were not located at their expected locations on the gel. Extensive analysis of the 11 corresponding isoforms showed that tryptic peptides from spot 23 matched exclusively with the central part of the protein, while spots 24 to 26 and 30 to 36 matched the N- and C-terminal extremities of the full-length 140-kDa precursor, respectively. In other cases, both the mature protein and cleavage products were identified. Other cases of discrepancy might correspond to computational errors in the identification of the true N-terminal boundary of the protein (i.e., the bona fide initiation codon). The same problem was noticed for proteins with functional attributes. According to both the areas and intensities of silver-stained spots, we can assume that some of the ORFan-encoded proteins are strongly expressed (spots 47 to 51, MIMI_L725; spots 52 to 53, MIMI_L688; and spot 61, MIMI_R653). Only four proteins not detected by LC-MS/MS were identified through 2D electrophoresis coupled with MALDI-TOF-MS.

FIG. 3.

Glycosylation pattern of Acanthamoeba polyphaga mimivirus proteins detected by fluorescence assay. Mimivirus proteins were separated by 10% SDS-PAGE (A) or by 2D electrophoresis using 7-cm IPG strips (pI 3 to 10) (B), and glycosylated proteins were revealed with GlycoProfile III. Lane 1, standard glycosylation markers; lane 2, mimivirus.

Antigenic properties of ORFan-encoded mimivirus proteins.

To investigate the antigenicity of mimivirus proteins, immunoblots were performed on samples resolved by 2D-PAGE, using sera from mice immunized with the whole particle (Fig. 4). Interestingly, two ORFan-encoded proteins, namely, MIMI_L724 (218 amino acids) and MIMI_L725 (224 amino acids), were specifically recognized, while no protein was recognized by preimmune sera. Both proteins are rich in cysteine residues and might be involved in a cross-linked structure accessible at the periphery of the particle. Immunogold labeling using an anti-MIMI_L725 monoclonal antibody further revealed gold particles surrounding the capsid core structure of mimivirus, confirming the exposure of the corresponding epitope to the surface (Fig. 5).

FIG. 4.

Western blot analysis of mimivirus proteins. Mimivirus proteins were separated by 2D electrophoresis using 7-cm IPG strips (pI 3 to 10) and transferred to nitrocellulose membranes before incubation with serum from an immunized mouse as described in Materials and Methods.

FIG. 5.

Localization of Mimi_L725-encoded epitope. Ultrathin sections of mimivirus-infected amoebae were immunogold labeled with a monoclonal antibody raised against the MIMI_L725-encoded protein followed by goat anti-mouse antibodies conjugated to 25-nm colloidal gold particles before immunoelectron microscopy analysis. Bar, 200 nm.

DISCUSSION

The proteomic analysis of mimivirus allowed us to identify 114 distinct proteins. Most of these proteins were identified through 1D electrophoresis coupled with LC-MS/MS, as this approach eliminated the methodological deficiencies encountered in separating low-abundance and hydrophobic proteins that are difficult to resolve by 2D electrophoresis (31). Completion of 2D gels coupled to MALDI-TOF-MS therefore provided additional and complementary information, as concluded from two recently published vaccinia virion proteome analyses (6, 48). First, for the majority of proteins identified from these gels, a good correlation between observed and theoretical pI and M_w values was observed. Analysis of 2D gels also revealed the existence of several isoforms due, in part, to protein glycosylation. Mimivirus contains six ORFs encoding putative glycosyltransferases and is thus, like Paramecium bursaria Chlorella virus (PBCV-1) (14, 46), likely to provide its own glycosylation machinery (33). The 2D gel profiles revealed cleaved proteins, a phenomenon found in many viruses and associated with synthesis and maturation steps (18, 27, 45). However, possible partial degradation of our samples due to contaminating proteases (although antiproteases were used) cannot be formally excluded, despite the reproducibility of our gel profiles.

This study also provided direct evidence that a large number of ORFan gene products contribute to the making of the mimivirus particle, with some of them acting as major components (based on the areas and intensities of the corresponding spots on 2D gels). Moreover, we demonstrated that some of these ORFan proteins are endowed with antigenic properties. Thus, proteins encoded by MIMI_L724 and MIMI_L725 were recognized by sera of mice experimentally infected with mimivirus. The antigenicity of MIMI_L724 as well as that of other ORFan proteins, encoded by MIMI_L330, MIMI_L442, and MIMI_L591, was also observed in a case of human infection (34). The surface exposure of the MIMI_L725 epitopes was further evidenced by immunogold staining with a monoclonal antibody raised against this protein. In the future, complementary experiments will be planned to localize other identified antigens. It is noteworthy that no eukaryotic proteins were identified in our samples, confirming their purity and demonstrating that eukaryotic proteins are not encapsidated within mimivirus particles. In contrast, proteomic analyses of vaccinia virus carried out by two distinct laboratories identified a significant proportion (>20%) of host proteins associated with purified intracellular mature virions (6, 17). Virus-associated host proteins include highly conserved proteins such as β-actin (6). This observation, associated with the fact that two amoebal genomes are available in the databases (10, 25), led us to discard a possible failure of identification. Genome replication and assembly of viruses often take place in specific cytoplasmic compartments, termed the viral factory (28), that may be instrumental in limiting the packaging of eukaryotic components within the viral particle. This hypothesis fits well with our own observation of mimivirus factories in infected amoebae (unpublished data).

Pending future studies of mimivirus physiology and future characterization of the virus-host interaction, the functions predicted for the proteins associated with the virion allow us to speculate on a possible molecular scenario for the early stage of the virus cycle. For instance, it is likely that the encapsidated lipolytic and proteolytic enzymes are part of the mechanism allowing the virus to gain access to the amoeba cytoplasm from its initial vacuolar location. The two topoisomerases (IA and IB) might then be involved in the ejection of the viral DNA from the particle and its injection into the cytoplasm, as well as allowing transcription to take place. These enzymes, which regulate the superhelical density of DNA (7), could indeed provide a driving force for DNA ejection into the bacterial cytoplasm by using the energy gained from changes of DNA topology (5). The DNA repair machinery components might then turn to action, allowing the viral DNA to be repaired prior to starting its transcription. Finally, the completeness of the transcription machinery components found in the viral particle is highly reminiscent of what has been described for poxviruses (15, 49), strongly suggesting that early transcription could take place in the host cytoplasm immediately upon mimivirus infection.

Altogether, these results contribute to the ongoing debate on the evolutionary origin of the gene content of large DNA viruses epitomized by mimivirus. Part of the community considers large viral genomes a “bag of genes” randomly accumulating unused laterally transferred genes (26). However, the recent release of two whole amoeba genomes (10, 25) does not favor this hypothesis (30). The fact that some of the mimivirus ORFans match unknown DNA sequences obtained from the Sargasso Sea suggests that mimivirus's uniqueness might well be related to a poor knowledge on viruses from the environment (13), which could partly explain the huge proportions of unmatched sequences in metagenomic studies (9). According to this view, the sequences of these genes might quickly lose any resemblance to their homologues under no functional constraint. Our finding of multiple ORFan products in viral particles rather suggests that ORFan proteins participate in a highly coordinated multimolecular process normally associated with tight evolutionary constraints. The unprecedented conservation of mimivirus promoters (42), as well as evidence for gene duplication generating well-conserved paralogues (20, 41), is also in favor of a well-ordered rather than chaotic model of genome evolution.

In conclusion, we believe that this work fuels the debate on the nature of viruses. Mimivirus was thought to be as complex and large as microorganisms of comparable genome size, such as rickettsiae or mycoplasmas (21). As a matter of fact, this proteomic analysis confirms that the virus's protein complexity is comparable to that of Rickettsia conorii (35). This observation, in association with the apparently common origin of the capsid protein of mimivirus and those of viruses infecting different hosts in the three domains of life, i.e., Eukarya, Bacteria, and Archaea (3), makes the question of the origin of viruses more relevant than ever (32). Finally, recent metagenomic studies indicated that up to 70% of randomly sampled DNA sequences from various environments have no database match, while 10% have similarities with viruses. Our data, with the finding of mimivirus relatives in the Sargasso Sea, suggest that environmental ORFans may indeed correspond to bona fide viral proteins whose functions and origins remain to be discovered.